Update app.py

app.py

@@ -37,7 +37,7 @@ DEFAULT_ZONE_FRAC = float(os.environ.get("GLMOCR_DEFAULT_ZONE_FRAC", "0.12"))
 MIN_CROP_HEIGHT = int(os.environ.get("GLMOCR_MIN_CROP_HEIGHT", "112"))
 MIN_CROP_PIXELS = int(os.environ.get("GLMOCR_MIN_CROP_PIXELS", "12544"))  # 112*112
 # Wider PDF bands for position-based extraction (capture account numbers etc. in top/bottom)
-PDF_HEADER_BAND_FRAC = float(os.environ.get("GLMOCR_PDF_HEADER_BAND", "0.15"))   # top 15%
+PDF_HEADER_BAND_FRAC = float(os.environ.get("GLMOCR_PDF_HEADER_BAND", "0.08"))   # top 8%
 PDF_FOOTER_BAND_FRAC = float(os.environ.get("GLMOCR_PDF_FOOTER_BAND", "0.85"))   # bottom 15%
 
 # Single shared parser to avoid "GLM-OCR initialized" per request and asyncio cleanup issues.
@@ -295,7 +295,8 @@ def run_ocr(uploaded_file):
                         ftr = extract_pdf_text_in_band(path, page_num, PDF_FOOTER_BAND_FRAC, 1.0)
                 if not (ftr and ftr.strip()):
                     ftr = ocr_zone(img_path, fs, 1.0)
-                if ftr and ftr.strip():
+                # Skip footer if its content already exists in page_md (avoids duplication)
+                if ftr and ftr.strip() and ftr.strip()[:80] not in page_md:
                     parts.append(ftr.strip())
 
             if parts: